基于2D单应矩阵约束的图像匹配方法

Image Matching Method Based on 2D Homography Matrix Constrain

针对目前普通图像匹配不精确的问题,用2D射影变换的单应矩阵约束估计算法和尺度不变特征变换(SIFT)算法结合,提出了一种适应性强,图像匹配精度高的图像匹配算法.首先对图像进行SIFT特征提取,在不同的尺度空间下提取特征角点,建立对应数据库.再对特征数据库进行初始化单应矩阵估计,剔除野值点后再用L-M(Levenberg-Marquardt)算法约束对单应矩阵继续估计,直到求出最优化的单应矩阵,最终根据优化单应矩阵生成内点数据库.

Aiming at the problem that ordinary image matching is not accurate, invariant feature transform （SIFT） algorithm is presented which is a strong adaptability and matching accuracy image matching algorithm by combining 2D projective transformation matrix estimation algorithm with scale invariant feature transform （SIFT） algorithm. First SIFT feature were extracted. Feature corners were extracted under different spatial scales, and then corresponding database were established. According to homography matrix, characteristic database matrix is estimated. After estimation the wild value points should be reoccupied and then applying L-M （Levenberg - Marquardt） constraints continue to the estimate the homography matrix, until finding out the optimization of it, finally according to the optimization of homography matrix the inner points can be generated.